Aducanumab: Less Than Meets The Eye

The beta-amyloid target in Alzheimer’s has become, to put it politely, somewhat tarnished over the past several years. AN-1792, bapineuzumab, solanezumab, gantenerumab, plus several BACE and gamma-secretase inhibitors/modulators, have combined to form a conga-line of clinical disappointments, undulating its way into oblivion. There are a few plucky survivors: Lilly has launched solanezumab into another pivotal Expedition in spite of widespread concerns that it will again be ‘too little, too late;’ Merck’s BACE inhibitor continues in Phase III, as they hold their breath against the fear that the safety issues that have derailed several BACEi peers might yet come to haunt them as well.

And there is aducanumab/BIIB037, Biogen’s (via a partnership with Neurimmune) mAB against beta amyloid plaque (as opposed to soluble AB). The unveiling of the data from a 166pt trial that was powered to address safety, not efficacy, has been a triumph of effectively selective disclosure. Last December 3, Biogen revealed at a Deutche Bank conclave that they had achieved success in a Phase Ib trial, albeit without much in the way of detail. They were not particularly concerned about disclosure standards, and indeed, continued to disseminate more clues about the trial results at other investor gatherings, such as at a Cowen meeting in Boston. Finally, on March 20, they presented more, albeit still partial, data at the AD/PD meeting in Nice. The resultant headlines were bullish, and by the end of the day,  Biogen had added another $10 billion or so to its market cap. To put it in perspective: Biogen’s valuation on December 2, the day before their AD data drip began, was $72 billion. By the end of the day on March 20, it had gained another $39 billion. While Biogen has a wide and impressive range of programs, and is anything but a ‘one-trick pony;’ in NIR’s estimation, most of that rise (perhaps all, since it is in spite of countervailing concerns regarding their EU Tecfidera franchise) can be attributed to this Alzheimer’s news.

Think about that for a moment: $39 billion in added value, attributable to a trial that enrolled a total of 166 patients, wherein the dose-cohort sizes were 30-32, and one of those cohorts (6mg/kg) did even not have its twelve month data reported (though it apparently has been completed). That dose was added later because of safety/tolerability concerns, and a high discontinuation rate, seen with the highest (10mg/kg) dose.

Beyond confirming the power of suggestion, what does the data from this Phase Ib study establish?
1) Target-engagement: It does appear that aducanumab successfully engages with the amyloid plaque target in a clear, dose-related fashion. This is not a minor point; it has been often wondered how many CNS drug failures reflected the inability to get the investigational drug where it needed to be, in sufficient quantity. It is pretty clear that this mAb does reduce plaque significantly, and while we have sometimes wondered whether this might not be a good thing (in theory, might this free up soluble amyloid for further toxic binding within the brain?) there was nothing in these results that pointed towards an iatrogenic effect.
2) Clinical Benefit: This is where the variance in ‘audience’ response broadens. For the most part, the media ‘drank the Kool-Aid’ and asked for seconds. But some Alzheimer’s luminaries who are not amyloid acolytes turned a firm thumbs down, albeit not for attribution. Most of the reported (and that is considerably less than what has been collected) datapoints, i.e. the mean changes in MMSE and CDR-SB scores, (adjusted for covariates like baseline scores and APOE4 carrier status) fell in line with a dose-related curve, the impact upon deterioration rates seeming to rise with higher dosing. But there are factors that mandate caution in taking these numbers at face value and embracing these findings prematurely:
a) There were four dose levels, 1mg/kg, 3mg/kg, 6mg/kg, and 10mg/kg, all infused on a monthly basis. The 6mg/kg dose, which was added after the study was underway, did not separate from placebo at 26 weeks on the CDR-SB or MMSE, unlike the other doses. The 54 week data, not yet reported, may yet show that 6mg takes its projected place between the 10mg/kg and 3mg/kg doses in terms of associated impact on cognition/function. But since it diverges at 26 weeks, that cannot be assumed. Given the high variance on these outcome measures, it also cannot be automatically assumed that it is the 6mg/kg dose’s absence of separation that is the anomaly, given the amount of ‘noise’ in the dataset.
b) There is a lot of data missing, and how that missing data was handled adds to the ‘noise.’ The highest dose (10mg/kg) group was also the one with the highest rate of discontinuations due to adverse events, 31%, whereas only 10% of the placebo group discontinued prematurely. Given that it was the APOE4 group that tended to account for the dropouts in the 10mg/kg group, not surprising since they were the most prone to ARIA (amyloid-related  imaging abnormality, indicating vasogenic edema), this was the dose-group with the highest frequency of earlier-in-the-study clinical data being carried forward to the 54 week readout (LOCF, last-observation-carried-forward). Since this is the more rapidly-deteriorating group,  using LOCF means that their final results were recorded as being better than they would have been had they actually completed the study.  This is in contrast to the placebo group, where E4 patients of course did not receive the mAb, and thus were not preferentially more likely to drop out due to side effects. Instead, their deteriorating course was incorporated into the 54 week data; these two factors in concert would bias the comparison towards enhancing the apparent treatment effect.
c) Speaking of missing data: Where were the results from the other neuropsychological testing components (the NTB and FRCT) with their far more ‘granular’ assessment of cognitive functions? The similarity between the results obtained from the MMSE and CDR-SB is less validating than it might seem: Those two instruments overlap quite a bit, their results tend to be highly correlated, adding to the need for confirmation from these other, more sophisticated psychometric batteries.  Biogen intends to release those results in July, but with all due respect to giving the investigators multiple days in the limelight, there is no excuse for not disclosing and discussing those results along with the MMSE and CDR-SB. After all, the valuation gained thus far by Biogen on the basis of these preliminary results is larger than the annual gross domestic product of some 97 individual countries (International Monetary Fund). The scale is huge, the stakes are high, and Biogen–who we consider to be an exceptionally well-run and effective company–has been  surprisingly cavalier and selective in its dissemination of these trial results.

3) Safety: Vasogenic edema is better understood now than when the first mAb trials reported their results, and it is now known that it tends to be transient. As is not surprising, there was a considerable incidence of vasogenic edema at higher doses. Most cases (92%) emerged in the first five months of treatment; 35% of patients developed subjectively salient symptoms–headache, blurred vision; 78% were described as ‘mild to moderate’, which means that 22% were more severe, some requiring hospitalization.In the APOE4 carrier population, 55% of E4 patients receiving the 10mg/kg dose developed edema. 43% of the E4 patients receiving the 6mg/kg dose did, the incidence was far lower (5%) amongst E4 patients on the 3mg/kg dose. It will require a far better understanding of aducanumab’s efficacy at various dose levels, likely differentiated by APOE genotype, to sort out the risk-benefit profile for aducanumab, and the degree of tightrope-walking that might eventually be necessary to achieve therapeutic benefit with acceptable tolerability. How acceptable this kind of risk will be for a treating physician may reflect physician temperament as much as anything: Some consider this profile quite ‘manageable’, but there are risk-averse prescribers who could find it discomfiting.

The bottom line is that this was a Phase Ib trial with tiny dose cohorts, particularly when broken down by genotype. The antibody gets where it needs to go, and does reduce amyloid plaque levels. Keeping in mind the caveats noted above regarding missing data and high intragroup variance, the trends shown here are in the direction that one would hope to see from a trial. But it neither proves nor disproves the case for the amyloid hypothesis or for aducanumab. It provides a highly provisional hint of an efficacy signal: It could be statistical noise, but if so, it is well-behaved, generally dose-congruent noise. The unreported neuropsych testing data will be useful in assessing the validity of these findings. Biogen, in keeping with the example set by its peers, is planning to go into Phase III later this year. That should not be taken as inferring that this trial provided anything approaching Phase IIb POC. They will roll the Phase III dice, rather than spend two+ years trying to replicate it with a Phase IIb. They are banking on the hope there is something more substantial than noise behind these data.

This entry was posted in BioFollies, Biotech, Muddled Media, Science and tagged , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s